Joint Cascade Face Detection and Alignment

ECCV (6), pp. 109-122, 2014.

被引用339|引用|浏览224|来源
EI
关键词
rotation invariantface alignmentsimple featureface detectionobject detection更多(2+)
微博一下
We present a face detector that establishes the new state-of-theart in terms of both accuracy and speed

摘要

We present a new state-of-the-art approach for face detection. The key idea is to combine face alignment with detection, observing that aligned face shapes provide better features for face classification. To make this combination more effective, our approach learns the two tasks jointly in the same cascade framework, by exploiting recent ...更多

代码

数据

0
简介
  • Face detection is one of the mostly studied problems in vision [31]. The seminal work of Viola and Jones [26] has established the two foundation principles for practical solutions: 1) boosted cascade structure; 2) simple features.
  • The authors' post classification is insufficient, because when high recall is expected, a cascade detector would return too many false positives and the SVM classifier would be slow too.
  • This results in new training and testing algorithms for joint cascade detection and alignment.
  • The authors present the first joint cascade face detection and alignment method.
重点内容
  • Face detection is one of the mostly studied problems in vision [31]
  • Many works are on multi-view face detection [10,17,27,7]
  • We present a face detector that establishes the new state-of-theart in terms of both accuracy and speed
  • To better exploit the alignment information, we propose a unified framework for cascade face detection and alignment
  • The complete training algorithm of our cascade face detection and alignment is summarized in Algorithm 3
结果
  • The authors firstly show that simple shape indexed features are effective for detection as well.
  • The authors illustrates the effectiveness of using face alignment for detection on the challenging FDDB dataset [9].
  • The face alignment based post classifier is brute force and too slow for a standard cascade detector when a high recall is desired.
  • The cascade detection is very fast because most negative image windows are rejected after evaluating only a few weak classifiers.
  • The general testing algorithm for cascade face detection and alignment for an image window x.
  • The authors propose to apply such features in detection as well, by making the learning of weak classifier Ci(x) in Eq (1) dependent on the face shape.
  • As shown in Section 3, any cascade shape indexed face alignment method [4,28,29,21] can be used for cascade detection.
  • Local learning of the tree structure: for each of the L facial point, a standard regression forest [2] is learnt to estimate the increment of this point, using the shape indexed pixel difference features [4].
  • The authors' testing algorithm for cascade face detection and alignment for an image window x.
  • To learn a mixed classification/regression decision tree, the authors use a similar strategy as in hough forest [6]: in the split test of each internal node, the authors randomly choose to either minimize the binary entropy for classification or the variance of the facial point increments for regression.
  • Training of cascade and joint face detection and alignment.
结论
  • 1: Input: all training samples {xi}, class labels {yi} 2: Input: ground truth shapes Si for positive samples, yi = 1 3: Output: all weak learners {CRtk}, classification thresholds {θkt } 4: set the initial face shapes S0i as random perturbations of the mean shapes in windows of xi
  • The complete training algorithm of the cascade face detection and alignment is summarized in Algorithm 3.
  • Afterwards, the authors train an alignment based SVM post classifier using the output windows in the training images, as described in Section 2.
  • The authors firstly compare the false positive rates with respect to the number of tested weak classifiers.
总结
  • Face detection is one of the mostly studied problems in vision [31]. The seminal work of Viola and Jones [26] has established the two foundation principles for practical solutions: 1) boosted cascade structure; 2) simple features.
  • The authors' post classification is insufficient, because when high recall is expected, a cascade detector would return too many false positives and the SVM classifier would be slow too.
  • This results in new training and testing algorithms for joint cascade detection and alignment.
  • The authors present the first joint cascade face detection and alignment method.
  • The authors firstly show that simple shape indexed features are effective for detection as well.
  • The authors illustrates the effectiveness of using face alignment for detection on the challenging FDDB dataset [9].
  • The face alignment based post classifier is brute force and too slow for a standard cascade detector when a high recall is desired.
  • The cascade detection is very fast because most negative image windows are rejected after evaluating only a few weak classifiers.
  • The general testing algorithm for cascade face detection and alignment for an image window x.
  • The authors propose to apply such features in detection as well, by making the learning of weak classifier Ci(x) in Eq (1) dependent on the face shape.
  • As shown in Section 3, any cascade shape indexed face alignment method [4,28,29,21] can be used for cascade detection.
  • Local learning of the tree structure: for each of the L facial point, a standard regression forest [2] is learnt to estimate the increment of this point, using the shape indexed pixel difference features [4].
  • The authors' testing algorithm for cascade face detection and alignment for an image window x.
  • To learn a mixed classification/regression decision tree, the authors use a similar strategy as in hough forest [6]: in the split test of each internal node, the authors randomly choose to either minimize the binary entropy for classification or the variance of the facial point increments for regression.
  • Training of cascade and joint face detection and alignment.
  • 1: Input: all training samples {xi}, class labels {yi} 2: Input: ground truth shapes Si for positive samples, yi = 1 3: Output: all weak learners {CRtk}, classification thresholds {θkt } 4: set the initial face shapes S0i as random perturbations of the mean shapes in windows of xi
  • The complete training algorithm of the cascade face detection and alignment is summarized in Algorithm 3.
  • Afterwards, the authors train an alignment based SVM post classifier using the output windows in the training images, as described in Section 2.
  • The authors firstly compare the false positive rates with respect to the number of tested weak classifiers.
表格
  • Table1: Notations in this paper category
Download tables as Excel
引用论文
  • Bourdev, L.D., Brandt, J.: Robust Object Detection via Soft Cascade. In: Computer Vision and Pattern Recognition, vol. 2, pp. 236–243 (2005)
    Google ScholarLocate open access versionFindings
  • Breiman, L.: Random Forests. Machine Learning 45, 5–32 (2001)
    Google ScholarLocate open access versionFindings
  • Brubaker, S.C., Wu, J., Sun, J., Mullin, M.D., Rehg, J.M.: On the Design of Cascades of Boosted Ensembles for Face Detection. IJCV 77, 65–86 (2008)
    Google ScholarLocate open access versionFindings
  • Cao, X., Wei, Y., Wen, F., Sun, J.: Face Alignment by Explicit Shape Regression. In: Computer Vision and Pattern Recognition (2012)
    Google ScholarLocate open access versionFindings
  • Felzenszwalb, P.F., Girshick, R.B., McAllester, D.A., Ramanan, D.: Object Detection with Discriminatively Trained Part-Based Models. IEEE Transactions on Pattern Analysis and Machine Intelligence 32, 1627–1645 (2010)
    Google ScholarLocate open access versionFindings
  • Gall, J., Lempitsky, V.S.: Class-specific Hough forests for object detection. In: Computer Vision and Pattern Recognition, pp. 1022–1029 (2009)
    Google ScholarLocate open access versionFindings
  • Huang, C., Ai, H., Li, Y., Lao, S.: High-Performance Rotation Invariant Multiview Face Detection. IEEE Transactions on PAMI 29, 671–686 (2007)
    Google ScholarLocate open access versionFindings
  • Jain, V., Learned-Miller, E.: Online domain adaptation of a pre-trained cascade of classifiers. In: CVPR (2011)
    Google ScholarFindings
  • Jain, V., Learned-Miller, E.: Fddb: A benchmark for face detection in unconstrained settings. Tech. Rep. UM-CS-2010-009, University of Massachusetts, Amherst (2010)
    Google ScholarFindings
  • Jones, M.J., Viola, P.: Fast Multi-view Face Detection. In: CVPR (2003)
    Google ScholarFindings
  • Kalal, Z., Matas, J., Mikolajczyk, K.: Weighted Sampling for Large-Scale Boosting. In: British Machine Vision Conference (2008)
    Google ScholarFindings
  • Koestinger, M., Wohlhart, P., Roth, P.M., Bischof, H.: Robust face detection by simple means. In: DAGM 2012 CVAW Workshop 13. Li, H., Lin, Z., Brandt, J., Shen, X., Hua, G.: Efficient boosted exemplar-based face detection. In: CVPR (2014)
    Google ScholarLocate open access versionFindings
  • 14. Li, H., Hua, G., Lin, Z., Brandt, J., Yang, J.: Probabilistic Elastic Part Model for Unsupervised Face Detector Adaptation. In: ICCV (2013)
    Google ScholarFindings
  • 15. Li, J., Zhang, Y.: Learning surf cascade for fast and accurate object detection. In: CVPR (2013)
    Google ScholarFindings
  • 16. Li, J., Wang, T., Zhang, Y.: Face detection using SURF cascade. In: International Conference on Computer Vision (2011)
    Google ScholarLocate open access versionFindings
  • 17. Li, S.Z., Zhu, L., Zhang, Z., Blake, A., Zhang, H., Shum, H.: Statistical Learning of Multi-view Face Detection. In: Heyden, A., Sparr, G., Nielsen, M., Johansen, P. (eds.) ECCV 2002, Part IV. LNCS, vol. 2353, pp. 67–81. Springer, Heidelberg (2002)
    Google ScholarLocate open access versionFindings
  • 18. Mikolajczyk, K., Schmid, C., Zisserman, A.: Human detection based on a probabilistic assembly of robust part detectors. In: Pajdla, T., Matas, J(G.) (eds.) ECCV 2004. LNCS, vol. 3021, pp. 69–82. Springer, Heidelberg (2004)
    Google ScholarLocate open access versionFindings
  • 19. Dollar, P., Welinder, P., Perona, P.: Cascaded pose regression. In: CVPR (2010)
    Google ScholarFindings
  • 20. Pham, M.T., Gao, Y., Hoang, V.D.D., Cham, T.J.: Fast polygonal integration and its application in extending haar-like features to improve object detection. In: Computer Vision and Pattern Recognition, pp. 942–949 (2010)
    Google ScholarLocate open access versionFindings
  • 21. Ren, S., Cao, X., Wei, Y., Sun, J.: Face Alignment at 3000 FPS via Regressing Local Binary Features. In: Computer Vision and Pattern Recognition (2014)
    Google ScholarLocate open access versionFindings
  • 22. Schneiderman, H., Kanade, T.: Probabilistic Modeling of Local Appearance and Spatial Relationships for Object Recognition. In: CVPR, pp. 45–51 (1998)
    Google ScholarFindings
  • 23. Segui, S., Drozdzal, M., Radeva, P., Vitri, J.: An integrated approach to contextual face detection. In: ICPRAM (2012)
    Google ScholarFindings
  • 24. Shen, X., Lin, Z., Brandt, J., Wu, Y.: Detecting and Aligning Faces by Image Retrieval. In: Computer Vision and Pattern Recognition (2013)
    Google ScholarLocate open access versionFindings
  • 25. Venkatesh, B.S., Marcel, S.: Fast bounding box estimation based face detection. In: ECCV Workshop on Face Detection (2010)
    Google ScholarLocate open access versionFindings
  • 26. Viola, P.A., Jones, M.J.: Rapid Object Detection using a Boosted Cascade of Simple Features. In: Computer Vision and Pattern Recognition, pp. 511–518 (2001)
    Google ScholarLocate open access versionFindings
  • 27. Wu, B., Ai, H., Huang, C., Lao, S.: Fast Rotation Invariant Multi-View Face Detection Based on Real Adaboost. In: ICAFGR, pp. 79–84 (2004)
    Google ScholarFindings
  • 28. Xiong, X., DelaTorre, F.: Supervised Descent Method and its Applications to Face Alignment. In: Computer Vision and Pattern Recognition (2013)
    Google ScholarLocate open access versionFindings
  • 29. Sun, Y., Wang, X., Tang, X.: Deep convolutional network cascade for facial point detection. In: Computer Vision and Pattern Recognition (2013)
    Google ScholarLocate open access versionFindings
  • 30. Yan, J., Lei, Z., Wen, L., Li, S.Z.: The fastest deformable part model for object detection. In: CVPR (2014)
    Google ScholarFindings
  • 31. Zhang, C., Zhang, Z.: A Survey of Recent Advances in Face Detection (2010)
    Google ScholarFindings
  • 32. Zhu, X., Ramanan, D.: Face detection, pose estimation and landmark localization in the wild. In: Computer Vision and Pattern Recognition (2012)
    Google ScholarLocate open access versionFindings
您的评分 :
0

 

标签
评论